A Generic Algorithm for Position-Orientation Estimation With Microscopic Vision

Position and orientation (PO) estimation with microscopic vision is essential for various micromanipulation tasks. Herein, to improve the accuracy and flexibility of PO estimation, a generic algorithm is proposed based on discriminative correlation filter (DCF), in which a position-estimator and an orientation-estimator are combined into one framework with the developed mutual correction mechanism. The extraction of spectral features is utilized to decouple the rotation and translation transformations of target. And DCF is employed to rapidly estimate the orientation. In addition, the continuous convolution operator is implemented to obtain subgrid resolution in the position-estimator. At last, both the stability and the accuracy of PO estimation are verified. The noise fluctuation in response distribution is effectively restrained to improve the robustness by the mutual correction mechanism. And the introduction of continuous convolution operation can improve the accuracy of position estimation too. The position error of
$\sim 1.00~\mu \text{m}$
and the orientation error of ~0.20° are achieved. The comparison and application experiments validate the comprehensive performance of algorithm. Its advantages include the high accuracy, the strong robustness to rotating influence, the universality for various microfeatures.